The Pareto front in average-cost MDPs has a geometric structure—it's a convex polytope where vertices are deterministic policies and edges are convex combinations of them, enabling exact computation without scalarization.
This paper shows how to find the exact Pareto front (all optimal trade-offs between competing objectives) in average-cost multi-objective decision problems. The front forms a piecewise-linear surface where each corner represents a specific policy, and you can compute solutions without solving individual optimization problems.